The Infrastructure Sovereignty Shift: Why Cerebras’ IPO Range Hike Resets the AI Valuation Ceiling
The surge in demand for Cerebras Systems’ IPO—evidenced by the upward revision of its price range this weekend—marks a permanent market re-rating of AI valuations, where 'sovereign compute' infrastructure is now commanding a higher valuation premium than software-centric AI lab models.
The Infrastructure Premium
On May 10–11, 2026, Cerebras Systems significantly revised its IPO terms, increasing its price range from $115–$125 to $150–$160 per share and boosting the share count to 30 million. With reports of the book being over 20 times oversubscribed, the market is pricing Cerebras at an implied fully-diluted valuation approaching $48.8 billion.
This is not a traditional semiconductor valuation. At the high end of the revised range, Cerebras commands an Enterprise Value-to-trailing 2025 revenue multiple of approximately 90x–96x. For context, this is a massive premium over even industry leaders like Nvidia, which typically trade at a fraction of that multiple. The market is not valuing Cerebras on current-year earnings or even standard semiconductor growth; it is valuing the company as a locked-in, multi-year infrastructure utility.
The 'Sovereign Compute' Thesis
The primary catalyst for this valuation is the $24.6 billion in Remaining Performance Obligations (RPO) disclosed as of December 31, 2025. This backlog is largely anchored by a Master Relationship Agreement (MRA) with OpenAI, which includes commitments for 750MW of inference compute capacity through 2028, supported by a $1.0 billion working capital loan from the AI lab.
This is the "sovereign compute" thesis in action: by binding hyperscalers like AWS—which entered a strategic collaboration term sheet in March 2026—and frontier labs like OpenAI into multi-year, capacity-guaranteed contracts, Cerebras has effectively de-risked its revenue profile. In an era where AI compute remains the primary bottleneck for scaling intelligence, investors are signaling that they prefer the predictable, high-margin, contract-backed business model of hardware infrastructure over the more volatile, model-governance-dependent models of pure software AI labs.
The Governance Divergence
This valuation reset is unfolding against a stark regulatory backdrop. While hardware players like Cerebras face geopolitical supply chain and export control risks, software-centric labs like OpenAI are increasingly burdened by a different set of hazards: lifecycle compliance, safety, and transparency obligations under frameworks like the EU AI Act. The market is increasingly discounting the latter due to this "governance overhang"—a dynamic clearly illustrated by the persistent pressure on OpenAI's IPO timeline amidst the ongoing Musk v. Altman trial.
The Counter-Argument
The obvious risk to this "compute-first" valuation model is its extreme customer and supply chain concentration. With future revenue heavily weighted toward a limited number of hyperscalers and a single-source manufacturing dependency on TSMC, Cerebras is essentially betting its entire existence on the continued dominance of its specific wafer-scale architecture. Should the AI scaling laws hit a plateau, or should hyperscalers shift their architectural preferences, this $48.8 billion valuation will face a rapid, painful reassessment.